The bandgap of III-nitride materials, including (Al, Ga, In)—N and their alloys, extends from the very narrow gap of InN (0.7 eV) to the very wide gap of AlN (6.2 eV), making them highly suitable for optoelectronic applications such as light emitting diodes (LEDs), laser diodes, optical modulators, and detectors over a wide spectral range extending from the near infrared to the deep ultraviolet. Visible light LEDs and lasers can be obtained using InGaN in the active layers, while ultraviolet (UV) LEDs and lasers require the larger bandgap of AlGaN.
Visible spectrum LEDs based on InGaN and AlInGaP systems have reached maturity and are now in mass production. However, the development of UV LEDs is still hampered by a number of difficulties involving basic material properties of AlGaN alloys, especially those with high Al content. Compared to LEDs in the visible spectral range with external quantum efficiency (EQE, the ratio of extracted photons to injected electron-hole pairs) of more than 50%, deep UV LEDs, such as those emitting below 300 nm, have an EQE of only up to 1%.
UV LEDs with emission wavelengths in the range of 230-350 nm are expected to find a wide range of applications, most of which are based on the interaction between UV radiation and biological material [Khan et al., 2008]. Typical applications include surface sterilization, water purification, medical devices and biochemistry, light sources for ultra-high density optical recording, white lighting, fluorescence analysis, sensing, and zero-emission automobiles. Although under extensive research for many years, UV LEDs, especially those emitting below 300 nm, remain extremely inefficient when compared to their blue and green counterparts. For example, Hirayama et al. recently reported 10.5 mW single-chip LED operation at 282 nm and peak EQE of 1.2% [Hirayama et al., 2009].
The growth of III-nitrides onto the c-plane sapphire is well-established. However, III-nitride material grown on c-plane sapphire suffers from the presence of polarization fields due to the polar nature of crystal bonds, which lead to energy band bending and reduction of recombination efficiency in quantum heterostructures due to physical separation of electron-hole wave functions, commonly known as the Quantum Confined Stark Effect (QCSE). Due to lattice mismatch, III-nitride materials grown on sapphire suffer from a high density of defects such as dislocations and inversion domains. A number of methods have been developed to obtain high quality single crystal material for device applications, including optimization of the nucleation process and choice of buffer layers to accommodate lattice mismatch. Alternative substrates, such as (111) Si, lithium aluminum oxide (LiAlO3) and silicon carbide (SiC) of various crystallographic planes also have been used for certain applications. However, native GaN and AlN substrates are still under development and remain prohibitively expensive.
Poor current spreading has been one of the major stumbling blocks to obtaining high efficiency deep UV LEDs, due to difficulties in achieving highly conductive yet sufficiently thick n-type AlGaN bottom cladding layers with high Al content. In 2004, Adivarahan et al. proposed a “micro-pixel” LED. The device consists of a 10×10 micro-pixel LED array, with each pixel being a circular mesa of diameter 26 μm. The total physical dimension of the device is 500 μm×500 μm. Since the lateral distance for electron migration before its recombination with a hole is significantly reduced using such geometry, the differential resistance of the device is lowered to 9.8Ω, as compared to standard square geometry LEDs based on the same epitaxial layers with differential resistances from 40 to 14.4Ω [Adivarahan et al., 2004]. Also in 2004, Kim et al. investigated the trade-off between mesa size and output power of circular-geometry deep UV LEDs, and found that without obtaining more conductive n-type and p-type AlGaN cladding layers, the optimized diameter for circular-disk deep UV LED is limited to about 250 μm [Kim et al., 2004]. Rather than making micro-pixel arrays or simply reducing mesa sizes to improve current spreading in the n-AlGaN cladding layer in traditional mesa-etched LED structure, various research groups have also employed the laser lift-off (LLO) technique for the deep UV. In a vertical structure LED, current spreading is much more efficient since metal contacts can be made vertically on both sides of the LED. In 2006, Zhou et al. at Philips Lumileds reported a vertical injection thin-film deep UV LED based on AlGaN/AlGaN quantum well structures emitting at 280 nm and 325 nm. The authors inserted a pure GaN layer in the epitaxial structure as the lift-off sacrificial layer. The device has dimensions of 700×700 μm2, and under 700 mA CW driving current, emits 160 μW at 280 nm and 3.1 mW at 325 nm. In addition to vertical device design, the authors applied a surface roughening process to the surface of the n-layer after lift-off. The roughening technique further increased the optical output power of the 280 nm LED to 0.74 mW (4.6 times improvement) and the 325 nm LED to 8 mW (2.5 times improvement) [Zhou et al., 2006]. In the same year, Kawasaki et al. demonstrated a vertical structure deep UV LED emitting at 322 nm using the LLO technique as well. However, the emission was rather weak and not a single-peak, possibly due to damage of the epitaxial layers during the LLO process [Kawasaki et al., 2006]. The development of LLO vertical thin-film deep UV LED was marked by the demonstration of high power 280 nm LED by the Nitek company in 2009. Nitek reported nearly 5.5 mW operation under 25 A/cm2 CW drive current on a 1×1 mm2 vertical structure deep UV LED, with lifetime exceeding 2000 hours [Adivarahan et al., 2009].
Molecular Beam Epitaxy (MBE) has recently been used to develop deep UV LEDs [Nikishin et al., 2008]. However, the output power of these devices is still low compared to those grown by MOCVD. This can partly be attributed to the slow growth rate of epitaxial film grown by MBE, which is thus incapable of producing very thick AlN templates that reduce dislocation density. However, with the advent of thick HYPE-grown AlN templates and free-standing AlN substrates now available from various suppliers, MBE as the production tool solely for the deposition of “LED layers” (i.e. only the n-type, p-type and active layers) may find use as an effective method.
A further problem for UV LEDs is the relative absence of band structure potential fluctuations in AlGaN crystalline material. This is due to the nearly identical size of Al and Ga atoms. This is in contrast to visible light LEDs, where the difference in size of In and Ga atoms the InGaN alloys have a tendency for phase separation leading to band structure potential fluctuations. Thus, injected electrons and holes in the active region of an InGaN-based visible LED are forming excitons, which are localized in these potential fluctuations, which prevents them from diffusing and recombining non-radiatively in defects. Thus, the recombination occurs primarily radiatively leading to LEDs with very high internal quantum efficiency (IQE). Since this mechanism is absent in UV LEDs fabricated by standard methods of growing AlGaN, efforts have been made to introduce band structure potential fluctuations in AlGaN for UV LEDs. U.S. Pat. No. 7,498,182 discloses a UV LED made using an MBE technique that produces faceted growth of AlGaN. The faceted growth mechanism creates localized inhomogeneities. AlGaN alloys described in this invention show clearly compositional inhomogeneities using cathodoluminescence spectroscopy. However, their emission spectra exhibit two peaks, one weaker characterized as band edge emission and the other which is stronger and red-shifted is due to compositional inhomogeneities. This result is clearly described one of the group's papers [C. J. Collins et al., APL, Vol. 86, 031916(2005)].
Thus, there remains a need to improve the crystal growth conditions employed during the deposition of various epitaxial layers in fabricating a UV LED, to develop methods of introducing band structure potential fluctuations, and to develop device designs that eliminate cracking and enhance carrier injection.